Inference
EN

MiniMax M3.1 Flash vs Muse Wonder

Change models
MiniMax M3.1 FlashMiniMax
Muse WonderMeta
IndexAll categories, out of 100.
69.3 ±5.4
73.1 ±4.0 best
GapIndex points below the top.
−3.8 · no clear gap
Top
CoverageIndex weight that has results.
80%
85% best
SpeedOutput tokens per second.
100/s best
80/s
EloLMArena rating and rank.
—
1503
ReleasedMonth of release.
2026-10
2026-10
Input / 1MUS dollars per 1M input tokens.
$0.15 best
$4.00
Output / 1MUS dollars per 1M output tokens.
$0.60 best
$20.00
1K chats2K tokens in, 500 out, each.
$0.60 best
$18.00
1K documents20K tokens in, 1K out, each.
$3.60 best
$100.00
1K agent tasks200K tokens in, 20K out, each.
$42.00 best
$1,200.00
ContextMaximum tokens in one request.
1.05M
1.05M
AgenticMulti-step tasks with tools.
71.8
73.0 best
CodingCode writing and repair.
69.4
72.3 best
ReasoningLogic problems and puzzles.
67.9
73.1 best
MultimodalTasks with images and text.
58.1
—
KnowledgeFacts and expert knowledge.
68.9 best
68.6
MultilingualTasks in many languages.
—
—
InstructionTasks with strict prompt rules.
—
74.4
MathMath problems.
—
73.5

Benchmarks

22 shared
Only MiniMax M3.1 Flash: 14Only Muse Wonder: 40
MiniMax M3.1 Flash
Muse Wonder
Artificial Analysis Humanity's Last ExamKnowledge
48.0%17 / 222
50.1% best12 / 222
Artificial Analysis Long Context ReasoningReasoning
79.7%45 / 217
80.7% best37 / 217
Critical Physics TasksReasoning
22.3%25 / 214
28.3% best15 / 214
Artificial Analysis Omniscience AccuracyKnowledge
45.1% best37 / 164
38.0%59 / 164
GDPval-AA normalizedAgentic
47.9%31 / 133
63.5% best5 / 133
Vibe Code Bench v1.1Coding
73.9%45 / 123
83.1% best22 / 123
Artificial Analysis SciCodeCoding
53.4%39 / 114
60.3% best10 / 114
Legal Research BenchLegal
34.1%50 / 90
50.5% best7 / 90
Tax Agent BenchFinance
32.1%23 / 82
36.3% best15 / 82
LiveBench Data AnalysisReasoning
71.4%52 / 80
81.6% best5 / 80
LiveBench ReasoningReasoning
82.9%48 / 80
90.6% best11 / 80
Terminal-Bench 4.0Agentic
25.3%28 / 61
49.5% best11 / 61
DeepSWEAgentic
73.7%13 / 59
78.8% best2 / 59
Agent Arena steerabilityAgentic
−2.636 / 54
2.6 best22 / 54
Agent Arena task outcomeAgentic
−0.734 / 54
5.4 best13 / 54
LiveBench Coding 2.0Coding
83.9%26 / 52
86.2% best20 / 52
LiveBench DesignAgentic
74.0%31 / 52
75.1% best29 / 52
LiveBench ResearchAgentic
72.7%29 / 52
88.0% best2 / 52
LiveBench Long AgenticAgentic
34.7%29 / 51
53.2% best9 / 51
LiveBench Browser UseAgentic
92.0% best4 / 49
72.0%20 / 49